I think my square is pretty smart.  He can get from point a to point b all on his own.  Makes a brief wrong turn here and there, but that’s just because he remembers what he did last time that worked.  Maybe a little forward thinking about his next move can change that.  Not implementing full A* for this because, at this point, the river never turns back on itself, it always goes forward, so the path-finding can be pretty simple.